A High-Pixel-Density, Wide-DR Subretinal Prosthesis SoC with Power-Aware Light Adaptation and Charge-Recycling Local Dynamic Supply.

This paper presents a 400-pixel light-adaptive retinal prosthesis (LARP) system-on-chip (SoC) that aims to support visual perception restoration. The proposed LARP SoC incorporates the power-aware light adaptation (PALA) engine and exposure-controlled light sensor (ECLS) to extend the effective light-sensing dynamic range to 55.48dB within a constrained pixel size. By combining charge-recycling auto-level-sorting multi-supply (CAM) generation with local dynamic supply (LDS) selection, the proposed system optimizes stimulation energy, with simulations showing up to a 45% reduction in stimulation energy consumption compared with conventional fixed-supply stimulation systems. The proposed pixel-clustering architecture integrates compact 80-μm × 80-μm pixels and employs a local return-electrode scheme, with the aim of potentially enhancing visual resolution by confining stimulation currents locally. Ex vivo experiments using wild-type and rd10 mouse retinas demonstrate the functionality of the LARP SoC, showing that it successfully evokes neural responses comparable to those produced by a commercial stimulator.